1. Introduction

In order to rapidly penetrate the potential applications and markets for AlGaN-based deep-ultraviolet (DUV) light-emitting diodes (LEDs), which includes surface disinfection, detection of material, air/water purification, solar-blind communication, lithographic microfabrication, medical diagnostics and so on [1–3], it will be necessary to achieve higher LED output powers and minimize the total DUV LED fabrication cost. However, despite numerous development efforts, DUV LEDs continue to have much lower light output power (LOP) from single chips than the more widely used blue LEDs. The main reasons for this are as follows. The first one is related to the materials. The large lattice and thermal expansion coefficient mismatches between sapphire substrate and the epitaxial layers cause large numbers of dislocations, which leads to lower internal quantum efficiency (IQE), severe heat and poor device reliability [4,5]. Secondly, it is related to light extraction efficiency (LEE). The absorption of DUV photons at the top p-GaN layer [6], as well as strong transverse-magnetic (TM)-polarized emission significantly decrease the LEE [7]. Thirdly, current crowding is also a major unsolved problem. In DUV LEDs, in order to avoid the absorption of DUV-light before being extracted to air, a high Al fraction is required in n-AlGaN layer, which results in higher series resistance and much severer current crowding problems compared to that in GaN-based visible and near-UV LEDs [8].

The micron-scale LED (micro-LED) is acknowledged as a promising way to solve many problems. These structures exhibit advantages such as suppressed internal strain, enhanced LEE, more uniform current spreading, improved heat dissipation, high modulation bandwidth and so on because of their small configuration [9,10], and therefore have been a research area of considerable interest recently. However, some problems become more critical when the active region is reduced. So far, the performances of DUV LEDs are far below forecasts. There is still a lot of room for improvement.

Sidewalls in micro-LEDs play an important role in the extraction of light from the mesa structure. In micron-scale structures, most of the transversely propagating photons can reach the sidewalls without suffering absorption [11]. It's worth noting that the micro-ring LED has more sidewalls compared to the micro-circular LED, and the ring-shaped mesa including inner and outer sidewalls are coated with a reflective Al mirror, which redirects the propagation vector of photons towards the emissive interface, thereby enhancing the efficiency of light extraction [12]. Furthermore, the micro-ring LED has weaker self-heating effect compared with the micro-circular LED. On one hand, the junction temperature of the ring-shape device is remarkably lower than that of the circular-shape device at the same current density because of its smaller size. The phenomenon of low junction temperature is related with the low average internal electrical resistance of ring-shape device caused by the short length of average current path between p- and n-electrodes, resulting in weaker self-heating [13,14]. On the other hand, ring-shape device has a larger sidewall area relative to circular-shape device, which is more conducive to heat dissipation and also reduces the heat generated by internal absorption of light due to the improved the extraction of light [15,16].

In this work, we succeeded in fabricating various AlGaN-based deep-ultraviolet (DUV) micro-ring LEDs, including varied ring sizes and the p-GaN-removed region widths at the outer micro-ring, respectively (PRM LEDs). The light output power density (LOPD) characteristics are investigated in detail. It is found that compared to micro-circular LED, micro-ring LED does have advantages in LEE. Besides, within a certain size range, smaller micro-ring LED can withstand higher current density and has higher LOPD. And through two-dimensional (2D)-finite difference time domain (FDTD) simulations, it is verified that our structure is indeed beneficial for improving LEE. However, owing to the more defects on the external and internal sidewalls, the micro-ring LED has a larger surface-to-volume ratio when the size continues to decline, which may lead to lower LOPD. Moreover, the influence of the the p-GaN-removed region and the corresponding effective light emission region on the electrical and optical performances of DUV micro-ring LEDs is explored. The PRM LED chip with a 90 µm external diameter, a 22 µm internal diameter, and an 8 µm p-GaN-removed region width is achieved, showing a maximum LOPD of 53.36 W/cm² at 650 A/cm² and an external quantum efficiency (EQE) of 6.17% at 5 A/cm². This work provides a path for advancing the application of DUV LEDs.

2. Experimental

In this experiment, the DUV LED epitaxial wafer used was commercial epitaxial wafer. It consisted of a 3 µm-thick n-Al_0.63Ga_0.36N layer, 4 periods of Al_0.42Ga_0.58N /Al_0.52Ga_0.48N multiple quantum wells (MQWs) layers, a 12 nm-thick p-Al_0.8Ga_0.2N electron blocking layer (EBL), a 60 nm-thick p-Al_0.65Ga_0.35N layer, and a 50 nm-thick p-GaN layer. The doping concentration of the n-Al_0.63Ga_0.36N and p-GaN layers were approximately 5 × 10¹⁸ and 2 × 10²¹cm^-3, respectively. Here, the PRM LEDs were fabricated into chips using flip-chip techniques. As depicted in Fig. 1(a), the mainly fabrication process was as follows: first of all, etched chip trenches, as shown in Fig. S1 (Supplement 1). After that, a part of the top p-GaN was etched to form micron mesa by standard photolithography and inductively coupled plasma (ICP) etching technologies, and the etching depth was very shallow, within nanometer range. Then, in order to form PRM LED arrays, the epitaxial wafer continued to be etched from the p-GaN layer to the n-AlGaN layer by two-step ICP etching technology. The PRM LED arrays are shown in Fig. S2. After a series of treatments by chemical agent on the exposed the n-AlGaN, V/Al/Ni/Au (15 /230 /20 /50 nm) metal layers were deposited around the ring-shape mesas as the common n-electrode. The rapid thermal annealing (RTA) at 800 °C for 90 s in N₂ ambient was employed to form the Ohmic n-contact. The PRM LED arrays with n-electrode are shown in Fig. S3. The p-electrodes of Ni/Au (5 /5 nm) metal layers were deposited on the p-GaN layer and RTA at 550 °C for 3 min in O₂ ambient to form the Ohmic p-contact, as shown in Fig. S4. Next, SiO₂/SiN_x (500 /20 nm) passivation layers were deposited by plasma-enhanced chemical vapour deposition (PECVD) and SiO₂/SiN_x apertures on each ring-shape mesas were then defined by buffered-oxide-etch (BOE)-based wet etching. The PRM LED arrays with passivation layers are shown in Fig. S5. Finally, Al/Ni/Au (300 /40 /50 nm) metal layers were deposited on SiO₂/SiN_x passivation layers, n-electrode layer and p-electrode layer as a reflector. It helped that some photons transmitted to the top and side walls are effectively reflected back to the substrate side and probably escape out [17].

 

Fig. 1. (a) The schematic diagrams of the fabrication process for the PRM LED. (b) The top-view SEM image of the PRM LED with an external diameter of 90 µm, an internal diameter of 28 µm, and a p-GaN-removed region width of 8 µm. (c)/(d) The larger cross-sectional SEM image of the PRM LED with an external diameter of 90 µm, an internal diameter of 28 µm, and a p-GaN-removed region width of 8 µm. (e) The AFM image of p-GaN layer mesa after micro-lithography.

Download Full Size | PDF

3. Results and discussion

The top-view as well as lager cross-sectional scanning electron microscopy (SEM) images of the PRM LED with an external diameter of 90 µm, an internal diameter of 28 µm, and a p-GaN-removed region width of 8 µm are shown in Figs. 1(b)–1(d). The height of the micro-ring-mesa is approximately 1.2 µm. Besides, it can be seen that the micro-ring-mesa of PRM LED is not a vertical structure, the inter side facet forms an angel of about 80° with respect to the bottom facet, and the outer side facet forms an angel of about 75° with respect to the bottom facet. Figure 1(d) shows the atomic force microscopy (AFM) image result for 50 × 50 µm² of the p-GaN layer mesa after micro-lithography, and the etching depth is 15-20 nm.

3.1 Study on the optoelectronic performance of different PRM LED sizes (reduced structural size in equal proportion)

In the case of current spreading, the low conductivity of Al-rich AlGaN can degrade the IQE of the DUV LEDs [18]. Optimizing the conductivity of the n-AlGaN is a way of ameliorating the current crowding situation. As the Al mole fraction in n-AlGaN increases, it may increase the contact resistivity and obtain higher rectifying behavior. V/Al-based metallizations have proven to provide lower contact resistivities at lower annealing temperatures for Al mole fractions higher than 0.4 [19]. Here, we choose V/Al/Ni/Au as n-electrode. In Fig. 2(a), the electrical characterization of the V/Al/Ni/Au contact on n-AlGaN layer is performed by current-voltage (I-V) measurements using transfer length measurement (TLM) pattern. The electrode width (W_c) is 100 µm, and the electrode intervals (L₁ to L₅) are 15, 20, 25, 30, and 35 µm, respectively. The extracted specific contact resistance is ∼2.37 × 10⁻³ Ω⋅cm² on the basis of curve fitting.

 

Fig. 2. (a) The change of extracted resistance with TLM pattern spacing. (b) The microscope images of the PRM-11, PRM-12, PRM-13, PRM-14 LEDs. (c) The LOPDs as a function of current densities for PRM-11, PRM-12, PRM-13, PRM-14 LEDs. (d) The 2D-FDTD simulation results of the cross-sectional electric field distribution for PRM-11 and PRM-13 LEDs. The distributions are shown for both the TE-mode and the TM-mode, respectively.

Download Full Size | PDF

The PRM LEDs with the external diameters of 130, 110, 90, and 70 µm, and the internal diameters of 42, 34, 28, and 22 µm [the proportion of hole area in entire mesa column (before is etched into ring) is 10%], respectively, and all the p-GaN-removed region widths of 8 µm, which are labeled as PRM-11, PRM-12, PRM-13, PRM-14, are first studied, as shown in Fig. 2(b). It can be seen that the morphologies of the nine micro-ring-mesas are almost the same, as highly uniform array. Figure 2(c) shows the LOPDs as a function of current densities for PRM-11, PRM-12, PRM-13, PRM-14 LEDs. The LOPDs of PRM-11, PRM-12, PRM-13, PRM-14 LEDs increase with the applied current density, and the values reach the maxima at 400, 500, 700, and 750 A/cm², respectively. The smaller LED withstands higher current density mainly owing to the better uniformity of the current spreading. For the PRM-11, PRM-12, PRM-13 LEDs, the smaller LED can provide higher LOPD at the same current density, which shows evident size effect, partly because LEE is improved. This explanation is also confirmed by the 2D-FDTD simulation. The perfectly matched layers (PML) for boundary condition are adopted in LEE simulation, which absorbs 100% DUV light beyond the calculated region. The refractive indices of materials are set as 2.6, 2.9, 2.16, 1.8 for AlGaN, GaN, AlN, sapphire, respectively. In addition, the material absorption in p-GaN and MQWs layers intensely restrict the LEE of DUV-LED. Their absorption coefficients are assumed to be 1 × 10³ and 1.7 × 10⁵cm^-1, respectively [20]. A series of electric-dipole sources emitting ∼280 nm light with TE/TM polarization are positioned at the MQWs layer. And the gaps between adjacent dipole sources are set as 0.1 µm, which is smaller than wavelength to avoid inhomogeneous interference fringes. The monitor is placed 1 µm above sapphire surface to calculate LEE, avoiding the affect caused by evanescent wave. Furthermore, in order to consider the current crowding effect due to the lateral carrier transport in the highly resistive n-AlGaN layer grown on insulating sapphire substrate, a weighting factor is introduced in the intensity of uniformly distributed each dipole source as a function J(x) = J(0)·exp(−x/L_s) following the current density, where x is the lateral distance of each dipole source from MQWs layer center. And current spreading length (L_s) value is set as 1.8 times of LED’ s lateral dimension [8,21]. The 2D-FDTD simulation is performed to investigate the effect of PRM-11 and PRM-13 LEDs on the LEE, as shown in Fig. 2(d). The TE-mode and TM-mode LEEs in PRM-11 LED are ∼5.613% and ∼0.034%, respectively. As for PRM-13 LED, the TE-mode and TM-mode LEEs are 6.293% and 0.050%, respectively, showing 1.12 times and 1.47 times that of PRM-11 LED. Smaller micron-ring-mesas have stronger waveguide effect to vertically confine photons, thus enhancing the light extraction [22]. However, under the same current density, the LOPD of PRM-14 LED is lower than that of PRM-13 LED due to the existence of more non-radiative recombination caused by etching damage. At 750 A/cm², PRM-13 LED reaches a maximum LOPD of 50.24 W/cm², which is higher than that of PRM-14 LED (50.13 W/cm²). Therefore, based on the result of PRM-13 LED, the size of the central hole is studied.

3.2 Study on the optoelectronic performance of different internal diameters of micro-ring LEDs at the same of external diameter

Figure 3(a) shows the microscope images of PRM LEDs with different internal diameters of 0, 12, 22, and 28 µm, respectively, which are labeled as PRM-21, PRM-22, PRM-23, PRM-13. Their external diameters are all 90 µm, and their p-GaN-removed region widths are all 8 µm. All mesa geometries lead to a saturation of the LOPD. The ring-LED saturates at a higher injection current relative to the circular-LED. The length of the current flow in the n-AlGaN layer under the ring-shaped mesa is shorter relative to the circular-shaped mesa, leading to better uniformity of the current spreading [23]. As the central hole size increases, LOPD increases at same current densities, mainly due to the influence of the suppression of light absorption loss and homogeneous injection current distribution [18]. The etched central hole continues to increase, exacerbating the negative impact of etching damage, thereby leading to a decrease in device performance. PRM-23 LED provides a maximum LOPD of up to 53.36 W/cm², which is 1.85 times that of the PRM-21 LED with a maximum LOPD of 28.92 W/cm². Compared to circular-shaped mesa geometry, the sidewall surface of the ring-shaped mesa geometry has been increased, which enhances the effective reflection of the photons from the side walls back to the substrate side and facilitates their escape. In order to better verify the improvement of LEE, the 2D-FDTD simulations for PRM-21 and PRM-23 LEDs are performed, as shown in Fig. 3(c). The TE-mode LEE is 5.497% and the TM-mode LEE is 0.011% for PRM-21 LED. As for PRM-23 LED, LEEs of the TE-mode and TM-mode are 6.083% and 0.058%, respectively, showing 1.11 times and 5.27 times those of PRM-21 LED. Further analysis of the optical features in this group of PRM LED chips is conducted, as shown in Fig. 3(d). As the current density increases, the peak wavelength and the full width at half maximum (FWHM) of the MQWs emission decrease, which are attributed to the decrease of the QCSE by Coulomb screening of the polarization field. However, the peak wavelength and the FWHM increase due to the self-heating effects in higher current density region. From 100 to 300 A/cm², PRM-21, PRM-22, PRM-23, PRM-13 LEDs present 0.7, 0.6, 0.4, 0.5 nm redshift, respectively. Since the red shift of the emission wavelength is mainly associated with junction-temperature rise [24], it is reasonable to deduce that, under the same current densities, PRM-13 LED suffers more severe self-heating effect resulting from more non-radiative recombination caused by etching damage relative to PRM-23 LED, which is consistent with the result of the LOPD. Figure 3(e) depicts the room temperature normalized electroluminescence (EL) spectra of PRM-23 LED driven under different current densities. It is noteworthy that the FWHM of the EL spectra are broad, which could be due to multiple causes. The fluctuations of carbon (C) and hydrogen (H) atoms before the EBL along with the consistency of alloy composition may contribute to the broadening of the EL spectrum due to carrier scattering and the presence of non-uniform potential wells. Besides, the fluctuations of materials such as aluminum (Al), gallium (Ga), and carbon (C) within the AlGaN layers or at interfaces due to kinetic adsorption processes also affect the spectrum broadening [25–30].

 

Fig. 3. (a) The microscope images of PRM-21, PRM-22, PRM-23, PRM-13 LEDs. (b) The LOPDs as a function of current densities for PRM-21, PRM-22, PRM-23, PRM-13 LEDs. (c) The 2D-FDTD simulation results of the cross-sectional electric field distribution in PRM-21 and PRM-23 LEDs. The distributions are shown for both the TE-mode and the TM-mode, respectively. (d) The wavelength and the FWHM as a function of the current densities for PRM-21, PRM-22, PRM-23, PRM-13 LEDs. (e) Normalized EL spectra of PRM-23 LED driven under different current densities.

Download Full Size | PDF

3.3 Study on the optoelectronic performance of different p-GaN-removed region widths at the outer micro-ring for PRM LEDs

Though the p-AlGaN is transparent in the DUV region, it is difficult to realize ohmic contact of p-AlGaN as Al composition increases even with high Mg doping because of the relatively high work function [31]. Therefore, p-GaN, which strongly absorbs UV light, is typically used as a p-contact layer in DUV LEDs. The selective removal of p-GaN and the formation of a reflective electrode on the unetched p-GaN are used to minimize the absorption and enhance the reflectivity. Here, the optoelectronic performance of different p-GaN-removed region at the outer micro-ring widths for PRM LEDs are explored. The PRM LEDs with p-GaN-removed region widths of 0, 4, 8, 11 µm, respectively, and all the external diameters of 90 µm, the internal diameters of 22 µm are displayed in Fig. 4(a), which are labeled as PRM-31, PRM-32, PRM-23, PRM-34. Figure 4(b) shows 2D-FDTD simulated LEEs results for PRM-31 and PRM-23 LEDs. Depending on the variation p-GaN-removed region width (from 0 to 8 µm), TE-mode LEE increases from 4.260% to 6.083%, whereas TM-mode LEE increases from 0.052% to 0.058%, indicating that the etched p-GaN layer decreases the absorption of photons. In Fig. 4(c), for PRM-31, PRM-32 and PRM-23 LEDs, the PRM LEDs with lager p-GaN-removed region width can be more tolerant with larger current density injection and also produce higher LOPD. In the contrary, PRM-34 LED shows a downward trend. The expansion of p-GaN-removed region has advantages and disadvantages. The increase of p-GaN-removed region results in a decrease in the current flowing to the sidewall, making the current more evenly distributed, thereby increases the withstand current density. At the same time, the reduction of the p-GaN layer decreases absorption of the DUV light, which leads to the enhancement of LOPD. However, the removal of excessive p-GaN region will result in the decrease of the effective light emission region and LOPD of the device. Figure 4(d) shows the I-V characteristics of PRM-31, PRM-32, PRM-23, PRM-34 LEDs, which have a threshold voltage about 5 V at forward bias. The EQE values can be calculated from the LOP measurement using expression (1) [32]. We also extracted the EQE values of these LEDs, and plotted in Fig. 4(f).

 

Fig. 4. (a) The microscope images of PRM-31, PRM-32, PRM-23, PRM-34 LEDs. (b) The 2D-FDTD simulation results of the cross-sectional electric field distribution in PRM-31 and PRM-23 LEDs. The distributions are shown for both the TE-mode and the TM-mode, respectively. (c) The LOPDs as a function of current densities for PRM-31, PRM-32, PRM-23, PRM-34 LEDs. (d) The I-V characteristics and (e) the EQE values for PRM-31, PRM-32, PRM-23, PRM-34 LEDs.

Download Full Size | PDF

4. Conclusion

In summary, different micro-ring LED arrays have been demonstrated, and are explored in three categories. The improvement in optical emission intensity can be attributed to the change of geometry structure. Moreover, the 2D-FDTD also verifies that the LEE of micro-ring LED is higher than that of micro-circular LED. Within a specific range, smaller micro-ring LED displays higher saturation current density and LOPD. However, too small micro-ring LED has a larger surface-to-volume ratio, which leads to lower LOPD due to the more defects-related non-radiative recombination on the micro-ring external and internal sidewalls. Furthermore, the value of the LOPD is synchronously influenced by p-GaN-removed region width as well as the corresponding effective light emission region, which possesses advantages and disadvantages, thereby necessitating an integrated consideration of the two. The optimized LED with enhanced optical and electrical properties will contribute to the development of AlGaN-based micro-LED arrays in DUV field.